Source: Apple is working on Apple Neural Engine, a dedicated chip to power AI on devices
Move would follow similar announcements from Qualcomm, Google — Offloading tasks to dedicated chip may improve iPhone battery — Apple Inc. got an early start in artificial intelligence software …
Context & Ripple Effects
In 2017, Bloomberg reported Apple was building a dedicated Apple Neural Engine to run AI tasks on devices, following similar dedicated-AI-silicon moves from Qualcomm and Google, with offloading framed as a way to spare the iPhone's battery. At the time it read as a component decision inside one product line.
Seen against the later coverage, it was the opening move of a full-stack strategy: Apple went on to acquire DarwinAI, whose tech makes AI systems smaller and faster (the 2024 acquisition), reportedly began designing its own AI chip for data center servers under Project ACDC (reported in May 2024), and by 2026 is expected to argue at WWDC that fifteen years of chip design gives it an edge running AI locally, including with a distilled Gemini model (per The Information).
First-order effects
- iPhone users are the immediate beneficiaries if the chip ships as described: moving AI tasks off the main processor onto a dedicated engine directly targets battery life, the constraint the report names explicitly.
- Qualcomm and Google, which had already announced comparable dedicated AI silicon, gain a third major rival validating their bet — and lose any first-mover differentiation in on-device AI hardware.
Second-order effects
- A dedicated neural engine pushes Apple's supplier base toward AI-specific IP blocks and manufacturing capacity, while pressuring other phone makers to match dedicated inference hardware rather than rely on general-purpose processors.
- Once inference lives on-device, Apple controls which AI features run locally versus in the cloud — leverage that later extends to its own servers via Project ACDC, letting it arbitrate the device-cloud split across its ecosystem.
Third-order effects
- If the pattern holds, smartphone competition shifts from app ecosystems alone to integrated AI stacks — silicon, models, and software designed together — which is exactly the argument Apple is positioned to make at WWDC with its distilled Gemini model.
- The device-to-data-center extension suggests the endgame is not one chip but a vertically owned compute path, where Apple runs AI locally on iPhones and on its own server silicon, reducing dependence on outside cloud providers for its AI features.
The trend: Consumer AI is consolidating around vertically integrated silicon strategies, with Apple's nine-year arc from the Neural Engine report to server-side Project ACDC showing dedicated AI compute becoming table stakes rather than differentiator.